Artificial Intelligence (AI) in clinical practice is receiving growing attention, given its potential impact on improving healthcare quality.
In particular, the integration of clinical guidelines in decision support systems and in the clinical workflow helps achieving treatment standardisation among clinicians.
A domain characterised by a scarcity of empirical data to guide clinicians' decisions is represented by psychotropic drugs prescriptions in pregnant women affected by depressive disorders. This is a particularly vulnerable group of patients for which randomised controlled trials to ascertain psychotropic drug's safety and efficacy are clearly unethical.
However, depressive symptoms during pregnancy are common (7-20%) and often require treatment, given the potentially dangerous consequences of untreated depressive symptoms for the mother and the child.
This highly interdisciplinary project aims at developing AI-based Virtual Doctors, i.e., mathematical models providing evidence-based treatment recommendations (in Selective Serotonin Reuptake Inhibitors administration during pregnancy) maximising efficacy and reducing side-effects.
Our unique team composition allows us to exploit:
1) Retrospective data spanning 5+ years (2015-2019+) from the Sapienza Centre for Prevention & Treatment of Women's Mental Health (Sant'Andrea Hospital), one of the largest centres in Italy, coordinated by the PI of this project.
2) Expert knowledge for all the relevant psychiatry domains, namely: psychopathology of mood disorders, psychopathology in pregnancy, pharmacotherapy, psychotherapy. Expertise on impact of traumatic life events on psychopathology will be covered by a post-doc position.
3) Long-term expertise in AI and Model Checking methods to design Virtual Doctors, particularly in the gynaecology domain (Tronci, coordinator of the EC FP7 PAEON project "Model Driven Computation of Treatments for Infertility Related Endocrinological Diseases").
[Max 5000 chars]
To date, AI techniques in healthcare have been mainly applied for the diagnosis, prognosis, treatment prediction, and the detection and monitoring of potential biomarkers.
Even more innovative possibilities concern the field of psychotropic drugs prescriptions in vulnerable populations such as pregnant women, where clinical practice suffers from the lack of randomised control trials, which are clearly unethical.
It is also worth noting that pregnancy does not exert a protective effect against mental illness and may even increase the risk if medication is discontinued.
This research project represents the first step towards explainable and trustworthy Artificial Intelligence [8] applied to the area of psychotropic drug prescription during pregnancy, a field characterised by the scarcity of empirical data to guide clinicians' decisions and in great need of standardisation.
We will develop AI-based Virtual Doctors as quantitative mathematical models that will precisely define the decision strategies followed by psychiatrists when treating pregnant women's depressive symptoms.
This goes way beyond classical clinical guidelines, because Virtual Doctors will encode the exact process to follow on the basis of patient data, both in terms of external factors (such as age) and clinical measurement (such as psychopathological symptoms, patients' temperament, etc.)
Our envisioned approach will radically improve the clinical management of this vulnerable group of patients, as the Virtual Doctors' treatment recommendations will perform explainable and trustworthy evidence-based reasoning aimed at maximising the efficacy and reducing drug side-effects.
This will represent an important advance in an area in which clinicians face the problem of finding a proper balance between the risks associated to drugs administration and the risks that untreated depressive symptoms pose for both the mother and the child.